\begin{table}[tb]
\centering
\caption{Effect of promotion on intervention ICS purchase in matched NGO/non-NGO villages\label{t:purchase_rates}}
\begin{threeparttable}
\begin{tabular}{lccc}
\toprule
& (1) & (2) & (3) \\ \cmidrule(lr){2-4}
& \multicolumn{3}{l}{\(\mathbbm{1} \left( \text{Purchased intervention ICS} \right)\)} \\
\midrule
\({TREATMENT}_j\) &  0.52*** &  0.45*** &  0.45*** \\
& (0.03) & (0.05) & (0.05) \\
\addlinespace
\({NGO}_j\) &     . & -0.03 & -0.04 \\
& (   .) & (0.03) & (0.03) \\
\addlinespace
\({TREATMENT}_j \times {NGO}_j\) &     . &  0.14** &  0.14** \\
& (   .) & (0.06) & (0.06) \\
\midrule
Control mean & 0 & 0 & 0  \\
District fixed-effects & Yes & Yes & Yes \\
Household-level controls & No & No & Yes \\
\(N\) & 943 & 943 & 943 \\
Adjusted \(R^{2}\) &  0.24 &  0.24 &  0.24   \\
\bottomrule
\end{tabular}
\begin{tablenotes}
\footnotesize
{\setlength\labelsep{0pt}
\item \textit{Notes}. The outcome variable is an indicator that equals 1 if household \(i\) in hamlet \(j\) purchased at least one of the two ICS promoted during the intervention. Column (1) presents aggregated results; results are disaggregated by NGO and non-NGO villages (as shown in Equation \ref{e:purchase_rates}) in columns (2) and (3). Baseline household-level controls for household size and number of children under five are included in column (3). Standard errors (in parentheses) are clustered at the hamlet level. \sym{*} \(p<0.10\), \sym{**} \(p<0.05\), \sym{***} \(p<0.01\).}
\end{tablenotes}
\end{threeparttable}
\end{table}
